In 2023, one of the biggest suppliers of rainforest carbon offsets was exposed: over 90% of their offsets were worthless. Despite industry objections to this claim, new research reveals that while some projects reduced deforestation, over-crediting was rife.
The carbon offsets formed part of REDD+, a voluntary climate change mitigation framework which encourages countries to avoid deforestation and forest degradation through providing financial incentives. Offsets are bought by organisations to ‘balance’ the carbon emissions from their operations.
The carbon credit industry raised concern over the validity of the research underpinning the exposé, citing that studies missed important local confounders – that is, variables that influence deforestation and project locations – that could’ve been incorporated in industry-led project evaluation.
The new study put these extra factors to the test. The researchers wanted to understand the level of unmeasured (ie unobserved) confounders that would close the gap between the research’s avoided deforestation estimates and those claimed by the early REDD+ projects – and whether this was plausible.
They found that, for many of the examined projects, these confounders would need to be implausibly strong – meaning that the estimates claimed by these REDD+ projects likely amounted to over-crediting. However, encouragingly, they found that projects did on average reduce deforestation, with some examples where projects deserved more credit than they received.
Lead researcher Dr Alejandro Guizar-Coutiño says:
“Unobserved confounders alone cannot ‘explain away’ the earlier findings of over-crediting in REDD+ projects; however, we did find true reductions in deforestation. Our findings have broader implications for nature market interventions – robust evaluation approaches, including rigorous sensitivity tests, are essential to ensuring projects are delivering on their intended outcomes and should be routinely used.”
Unobserved confounders are somewhat inevitable in observational designs, and are a key source of uncertainty when assessing the impact of nature market interventions and conservation efforts more generally. Alejandro adds: “The good news is, with the data and methods available today, we can get a reasonable approximation to keep our modelling assumptions in check.”
As part of the LEON project (Leveraging Earth Observation for Nature finance), the researchers are developing robust applications for quantifying biodiversity outcomes beyond carbon benefits to continue building on the findings from this study.
To read more about this research, published in Nature Ecology & Evolution, visit: https://www.nature.com/articles/s41559-026-03049-7
